felix642 / 1chipml Goto Github PK

Tests should follow this structure :

int main() {
    mat* input;
    mat* expected;
    
    test1(input, expected);

}

void test1(input, expected) {
    ...
}
...

When a test succeeds, it should return 0 and print "Success : test name with arguments"
When a test fails, it should return 1 and print "Fail : test name with arguments, expected expected value, but got real value

Add PCA example using Lanzcos and Jaocbi that can run on the atmega328p.

• Investigate method of gradient calculation (Finite Difference)
• Create pseudo-code for conjugate gradient
• Find Linear Minimization method

Detail changes to vision document

• Create uniform development environment for all team members

1. Define function with small number of variable and compute the min

2. On the edge ML model training (e.g., neural network on a relatively small problem)
   Implique le backpropagation, pas forcement reseaux de neuronnes
   Model avec hyperparametre pour voir ce qui donne les bonne prediction

3. Curve fitting from measurements (e.g. Simple short-term predictions/estimations)
   Recoit mesure sur microcontroleur (ex: Temperature)
   Curve fitting : donner valeur entre mesuration, ou apres mesuration

4 Convex optimization (e.g., for industry process optimization)
Trouver fct convex avec sens industriel
Pas trop de param sinon compliquer